My decision to dive into Genode came from this paper TCB comparison: An Overview of Microkernel Based Operating Systems (page 11). I immediately hoked into that low TCB number, And since then I’m on my journey here
After reading the Foundations book and literally spending a day to read through all the release notes available I’m still not sure what kernel to choose for my aims.
My goal is to run Sculpt on my homelab (Intel Gen 9) as a NAS (and potentially home router, etc.). And also another instance on my desktop PC (Intel Gen 10) to replace my ArchLinux setup.
From the release notes I got an impression that NOVA got the most attention throughout years while seL4 and HW were added later, with HW receiving the most attention in recent years. Other kernels supported by Genode seems to me are less maintained and thus less appealing.
Though the current state and readiness of each of them is still a mystery for me.
Of course Foundations book gives me many technical details about the kernels, as well as release notes about all the changes happened over 17 years. But I don’t have experience in OS development to understand their overall current state from those technical details.
As a newcomer I’d be happy to see some summary of what’s supported and what not by each kernel, like I can easily find for example on LineageOS pages per specific device, to have a quick decision with what to proceed. Either I didn’t find this summary page on Genode resources and forum or there is no such page
Judging from the efforts put into HW and how low TCB it gives it seems a way to go. But for some reason all 2025 SculptOS releases were provided with NOVA despite of limitations mentioned in the Foundations book. And if I got it right, Seoul (Vancouver) virtualization is supported only for NOVA, meaning no virtualization for HW (no VBox or Qemu?)…
So my question is rather simple:
What are the current limitations of each kernel + Sculpt per platform to be aware of if comparing to BSD or Linux out-of-the-box systems for the same platforms? Where can’t I get this insight?
Ok, 1 limitation of HW I’ve just accidentally found in a different topic:
HW kernel doesn’t support CPU mode change (power efficient/balanced/performance), right?
I believe currently NOVA is still the recommended kernel for a desktop configuration as it is the default provided with the pre-built SculptOS images, but it’s seeming like BaseHW is working very well too and may become the new standard soon. Totally agreed it would be great to have a clear listing of the limitations.
I hear there was a ‘wiki’ a few years back, though that was before I became a genodian so I didn’t have any experience with it. Word was, it was not easy for Genode Labs to maintain it.
I wonder if it would be possible to have something of a substitute here, which would be community-maintained : whenever a discussion topic here in Discourse ends up with a good response to a question, the response ends up in a “Master Topic” that is curated by a volunteer (no, I’m not volunteering ) and gets “sticky’ed” so as to always display at the top of the forum… Or if the response is more than a paragraph, the Master Wiki Topic links to the discussion instead of copy-pasting the answer to self.
People have little spare time these days… Especially programmers. But maybe there are non-programmers looking for ways to contribute, and who do have spare time ? This idea of a “community maintained wiki” might be for them. We would all strive to make things easy for said volunteer and provide help of course ^^
Virtualization on x86_64 is since 2019 (or short after [0]) in principle kernel independent. So, Vbox6 and Seoul should be runnable on all modern kernels. Well, why I say “should” or “principle” ? We don’t test all possible variations manually (man power, interest), so I think you should expect issues.
As Seoul maintainer I can only say, that it works for me on my Intel notebook and AMD desktop (Ryzen 5700) quite well for -nova-.
The slides and especially the backup slides (so they are not part of the video, but in the pdf) of the FOSDEM talk in 2025 may potentially answer to some degree your questions.
